Forecasting Elective Surgery Demand Using ARIMA-Machine Learning Hybrid Model

Abstract

Long wait times for elective surgery have not only caused patients to continue to live with inconvenience or pain but also creates frustrations and dissatisfaction with the local hospitals and healthcare systems. To deal with the increasing demand, hospitals need to be able to accurately predict the future demand to properly equip their facilities and the number of staff. In this paper, we propose various ARIMA-Machine Learning hybrid models to predict future elective surgery wait list demand. The goal of this paper is to improve the future demand predictions for hospital elective surgeries. We also compare our hybrid model to ARIMA and various Machine Learning/Deep Learning models, such as ANN, LSTM, and Random Forest. We found that ARIMA-ANN performed best with MAE of 0.26-0.76 and MSE of 0.13-1.05 with two-week-forward Urology, Orthopaedics and Gynecology elective surgery data.